NVIDIA currently dominates AI data center deployments with its highly successful A100 and H100 processors for artificial intelligence (AI) and high-performance computing (HPC) applications. But custom chips have become a global trend among large cloud service providers and emerging devices such as software-defined vehicles (SDV). According to Reuters, Nvidia is forming a new business unit to enter the custom chip market.
According to reports, the new business unit will be led by Vice President Dina McKinney, who has extensive experience working at AMD, Marvell and Qualcomm. The new unit is designed to serve the needs of a wide range of industries, including automotive, gaming consoles, data centers, telecommunications and other industries that can benefit from custom chip solutions. Although NVIDIA has not officially acknowledged the formation of the division, McKinney’s LinkedIn profile as vice president of silicon engineering shows that she is involved in the development of “cloud, 5G, gaming and automotive” chips, hinting at the broad scope of her alleged business unit.
Nine unofficial sources in the industry confirmed Reuters The division exists, but NVIDIA has been tight-lipped, only discussing announcements in 2022 about implementing its networking technology into third-party solutions.according to ReutersNVIDIA has begun discussions with leading technology companies such as Amazon, Meta, Microsoft, Google and OpenAI to investigate the potential of developing custom chips. This signals NVIDIA’s intention to expand its offerings beyond traditional off-the-shelf data center and gaming products to accommodate the growing trend of custom chip solutions.
While NVIDIA A100 and H100 processors are being used for AI and high-performance computing (HPC) instances, major cloud service providers (CSPs) such as Amazon Web Services, Google, and Microsoft are also improving their custom processors to meet specific AI needs. and general computing needs. This strategy allows them to cut costs and tailor the functionality and power consumption of the hardware to their specific needs. So while NVIDIA’s AI and HPC GPUs remain indispensable for many applications, more and more workloads are now running on custom-designed chips, meaning NVIDIA loses business opportunities. This shift to custom silicon solutions is widespread and the market is expanding rapidly. Essentially, NVIDIA doesn’t want to fight the custom chip trend, but rather join it.
Meanwhile, analysts are painting a bigger picture of possibilities. Jon Peddie Research, a well-known GPU industry observer, noted that they believe NVIDIA may be interested in not only meeting the needs of CSPs with data center products, but also in the huge capacity of the consumer market.
“NVIDIA has built a loyal fan base in the consumer market, which allows them to build a brand and develop more powerful processors that can then be used as computing accelerators,” said Jon Peddie, president of JPR. But the company has made a fortune in the deep-pocketed data center market, where mission-critical projects consider chip costs insignificant to overall goals. The consumer side provides NVIDIA with economies of scale so they can apply massive resources to develop Chips and the software infrastructure around those chips. It’s not just CUDA, it’s a huge library of software tools and libraries.”
NVIDIA tried to solve the problem for smartphones and tablets with its Tegra SoC back in the mid-2010s, without much success. However, the company has managed to achieve a position of supplying applications processors for the highly successful Nintendo Switch console, and is certainly looking to expand that business. The consumer business allows NVIDIA to design a chip and then sell it to a customer for years without changing its design, spreading the cost of developing the chip over millions.
“NVIDIA is certainly interested in expanding its presence in the console space, they are currently supplying the largest console vendors by volume and are calling on Microsoft and Sony every week to try to get back into the market,” Peddie said. “First NVIDIA is used in both the first-generation Xbox and PlayStation 3. But AMD’s APU has a cost-effective advantage, and NVIDIA hopes to match Grace. And because Windows runs on Arm, NVIDIA has the opportunity to challenge Microsoft. Sony’s customized operating system is for NVIDIA It won’t pose much of a challenge.”